Account Options

  1. Sign in
The old Google Groups will be going away soon, but your browser is incompatible with the new version.
Google Groups Home
« Groups Home
Message from discussion Hperthreading with SLURM- Issues
The group you are posting to is a Usenet group. Messages posted to this group will make your email address visible to anyone on the Internet.
Your reply message has not been sent.
Your post will appear after it is approved by moderators
 
From:
To:
Cc:
Followup To:
Add Cc | Add Followup-to | Edit Subject
Subject:
Validation:
For verification purposes please type the characters you see in the picture below or the numbers you hear by clicking the accessibility icon. Listen and type the numbers you hear
 
AMARNATH, Balachandar  
View profile  
 More options May 17 2012, 5:20 am
From: "AMARNATH, Balachandar" <BALACHANDAR.AMARN...@airbus.com>
Date: Thu, 17 May 2012 03:20:05 -0600
Local: Thurs, May 17 2012 5:20 am
Subject: [slurm-dev] Hperthreading with SLURM- Issues

Hello,

I am a new bie to SLURM.  First of all, thank you for the support you are rendering to the people.

I would like to share few configuration about my cluster configuration.

1.      The hyperthreading is disabled at SLURM level in my cluster. Each of the computing node has 4 cores. Hence, no more than 4 tasks can be submitted in one computing node (they are have 4 cores).
# COMPUTE NODES
NodeName=node[1-6] Procs=4 RealMemory=15360 Sockets=1 CoresPerSocket=4 ThreadsPerCore=1 State=UNKNOWN
PartitionName=batch Nodes=node[1-6] Default=YES MaxTime=INFINITE State=UP

2.      The /proc/cpuinfo in computing nodes has the following parameters in it

CPU     0       1       2       3       4       5       6       7
apicid  16      18      20      22      17      19      21      23
core id 0       1       2       3       0       1       2       3

From these values, what I interpreted is that both the logical CPUs 0 and 4 belong to core 0, and so on.

3.      The CPU allocation during parallel tests in one computing node (each has 4 cores) is as follows

Tasks   Thread  CPUs allocated  Corresponding core id
1       1       1       1
2       1       1,4     1,0
3       1       0,1,4   0,1,0  [ core 0 is overloaded?? I Should it be something like 0,1,2 here ? ]
4       1       0,1,4,5 0,1,0,1 [ core 0 and 1 is overloaded?? Should it be 0,1,3,4?]

I use the following bash file to submit the job through slurm,

#!/bin/sh
#SBATCH --exclusive
mpirun  -x OMP_NUM_THREADS=1 -np <TASKS>  `MYAPPLICATION`  'PARAMETERS`

I see a performance drop when going beyond 2 tasks and I attribute this drop to the above reason, that is, SLURM overloading same core with more than one task. If this statement is correct, how to come out of this situation? If not, can someone tell me where things are going wrong. When I directly run with 'mpirun' command on the computing node, the tasks are spanned across  individual cores unlike the above case.

With many thanks and regards
Balachandar

The information in this e-mail is confidential. The contents may not be disclosed or used by anyone other than the addressee. Access to this e-mail by anyone else is unauthorised.
If you are not the intended recipient, please notify Airbus immediately and delete this e-mail.
Airbus cannot accept any responsibility for the accuracy or completeness of this e-mail as it has been sent over public networks. If you have any concerns over the content of this message or its Accuracy or Integrity, please contact Airbus immediately.
All outgoing e-mails from Airbus are checked using regularly updated virus scanning software but you should take whatever measures you deem to be appropriate to ensure that this message and any attachments are virus free.


 
You must Sign in before you can post messages.
To post a message you must first join this group.
Please update your nickname on the subscription settings page before posting.
You do not have the permission required to post.